Real-Time Automated Response Analytics

Sneha GautamTravel

Share this Post

In today’s digital world, we are witnessing a rapid semantics overload due to which, it has become increasingly challenging for a business to process and respond to queries and concerns.

Consumers use various mediums like email, chatting mediums, and social media platforms to raise their concerns and expect real time resolution to their issues thus it becomes time-consuming and taxing for an entity who is working behind the scene to process and respond in-situ, to the real-time data that is being generated rapidly across these various platforms.

Henceforth, it becomes the responsibility of the organization to identify the problem and provide a solution to the consumer in real time, after following the adequate process.


In the current scenario where customers’ expectations are changing and organizations want to maintain a reputation in the industry where they are seen as a brand that respects the needs of its consumers, it has become imperative to automate the process of Real Time Response generation.

For a business entity to respond in real-time, it needs to develop a model that is capable of apprehending, filtering and analyzing data to make useful business decisions. The model should be capable of determining that a response is required and at the same time should also be able to intelligently identify what, when & how of the responses or reactions.


The fast-growing human-generated content, which includes text, email, speech voice recordings, videos, social media post, etc. are almost unstructured. And customers expect organizations to uncover the deep insights from these unstructured data and transform them into actionable insights, accelerating the speed of the outcome.

Developing a machine learning pipeline is generally a complicated process. Text analysis is about leveraging tools, techniques, and algorithms to process and understand linguistic data, which is usually in unstructured forms like text, speech, etc. Within this pipeline, we use the tested and well-experimented strategies, techniques and workflows to gain useful insights by leveraging our travel and hospitality expertise.

At IGT, we developed an NLP solution framework with the capabilities to mine the incoming query text, parse that text and do text preprocessing, extract relevant information and classify them into various entities to evaluate the intent/sentiment of the query.

Having all this in place, a tailor-made response is generated and shared with the concern.

The process starts at the data ingestion point which can be from multiple sources like email, chatbox or social media/forums.

The framework is capable of recognizing who the customer is by checking the information of the query sender against the customer database. This enables us to provide improved tailor-made responses for that customer based on his/her historical data regarding queries if it exists.

The incoming query text is ingested by the NLP solution framework where the text is analyzed based on text processing rules that the machine learning model has. Then, an appropriate response is generated and is provided to the customer instantaneously.

This framework has not only reduced human effort, but also has improved efficiency and accuracy. And because of a set framework in place, it has become easy to recognize the pattern and types of queries of the concern which has significantly also reduced the turnaround time.

With the help of this Machine Learning based automated response generation system, the task has become less taxing because of its capability to analyze and respond in real-time. Along with this, the instantaneous generation of responses with accuracy and efficiency has also lead to customer satisfaction, trust and loyalty.

Let’s understand with an example of a customer email to a leading US carrier about baggage query:

At the core of our system, whenever a customer sends an email to the carrier enquiring about his baggage, the system gets triggered and starts handling the incoming message, it does the required preprocessing on the text corpus and feeds it into the feedforward Neural Network (NLP Layer), which in turn predicts the most likely response and provides the information related to the baggage like its current location, why it’s being delayed, etc. to the customer.

The NLP framework also extracts the relevant information so that the response is curated to match the customer’s question. The system aims at understanding the customer email and then starts to action based on that understanding and convey meaningful information.

About the Authors:

Ambuj Mittal is a Data Science Engineer at IGT Solutions’ Travel Analytics group. As a Machine Learning practitioner, he loves challenging real-life travel problems which can be solved using the power of Data Science and ML. He is skilled in Python, Machine Learning, Deep Learning, and Data Engineering. He is involved in development of AI and ML solution for our Travel and Hospitality customers. For further information he can be reached at

Qurratulain Saleem is a Data Engineer at IGT Solutions. He is skilled in Python, Data Structures, Algorithms, Machine Learning and Deep Learning. He is a Statistics and Math enthusiast. He is a passionate programmer and like solving algorithm problems and philosophizing about AI, algorithms, machine learning and their effects on our world and real life problems. He can be reached at